You might have noticed that one thing was conspicuously absent in my two earlier articles about the likely completion of the Z9 generation and the possible start of the next: focus.
The current subject detect systems from all the camera makers are basically machine learning taught. This is essentially a narrow form of AI. You test and adjust against known conditions and situations, and build a model that is put into the camera that covers those cases.
Here's where I have problems with the current Z9 generation form of subject detection: it's missing a lot of subjects, conditions, and situations. Nikon has tuned that a bit (e.g. the addition of Birds), but not 100% successfully in my judgment. The current system recognizes some birds, sometimes.
Event photographers probably have the most capable focus system model at present: humans are recognized well, in many different contexts, and across diversity of skin (including makeup), costume, and how the subject is facing. Still, problems exist in the current focus recognition, often with multiple subjects being difficult for the camera to choose the right one from (which is why there's a button press to jump to "next recognized subject"). Silhouette and facing away aren't always handled well, either. I'm sure Nikon knows all that by now, and building new test situations to inform the current model wouldn't be difficult, so I'd expect improvement in the next generation.
Sports and wildlife is another story.
You might remember that Nikon published a Sports Setting Guide with the Z9 very early on. Look at the sports they have recommendations for: soccer, skating, ski jumping, skiing, hockey, gymnastics, track and field, aquatics, table tennis, BMX and skateboarding. All sports that were easily tested in Japan, particularly given part of the testing was at the Tokyo Olympics. American football, baseball, basketball, lacrosse, and many more sports were conspicuously missing. Moreover, I completely disagree with some of Nikon's suggestions, so I'd say the Z9 focus training for sports was not complete and not fully vetted. Helmets also become a real problem at times.
Wildlife is worse. Initially the Z9 generation was reliable only on dogs and most (but surprisingly not all) cats. Birds were added, though there are still birds the current system still has real problems with. I can list a whole host of animals that the current system will struggle to get right. And by "right" I mean "see body, identify head, find eyes," as that's the current learning priority being used. Leopards and cheetahs are hit-or-miss with the current system, as Z9 generation cameras seem to be fooled by spots on the sides of animals.
So here's why I left focus to a separate article: focus is probably the one new function/feature that got DSLR and original Z system users to move to the Z9 generation: the focus system is better in the Z9 generation and enough so that it provides both amateurs and pros many clear benefits. However, if I can clearly identify use cases where the system has problems, the question is has Nikon, and if so, have they done the next generation machine learning necessary to boost all of those cases up to the level the current camera gets right most of the time for humans, plus domestic dogs and cats?
I don't know what's possible with EXPEED7 in this respect. Nikon needs a larger, more nuanced, and potentially faster model in the camera. Will it fit and could the current digital hardware give it current or next level performance across more subjects and conditions? Maybe. Maybe it will take more digital horsepower.
So when I asked whether a Z9II was necessary, I left out a key component: augmenting and adjusting the focus system. Moreover, I don't know if Nikon has—particularly for wildlife in action in real environments—the test beds to train the system better.
Thus, I left out the focus system in my earlier articles because three more possibilities exist:
- Minimal — Current subject recognition and tracking has been retrained to a higher level.
- Modest — Nikon trained in scenarios and with subjects they neglected the first time around (plus did the Minimal work).
- Maximum — Nikon added "new sauce" to the technologies involved, plus did the Modest/Minimal work).
What would a "new sauce" be? Well, the one we (other than Canon users) are all waiting for is dual axis recognition. This is exactly the thing that made the D6 (with its larger, all cross-hatch sensors) so much better than the D5, and made the D6 the best focusing camera of any I've seen to date (at least in the central area).
Can Nikon repeat that D5->D6 gain in the Z9->Z9II changes? Certainly the technology exists that might achieve that. Moreover, that would be a generational pass-down tech that could trigger the next upgrade cycle for everyone.
